5 #include <linux/module.h>
7 #include <linux/genhd.h>
8 #include <linux/kdev_t.h>
9 #include <linux/kernel.h>
10 #include <linux/blkdev.h>
11 #include <linux/init.h>
12 #include <linux/spinlock.h>
13 #include <linux/proc_fs.h>
14 #include <linux/seq_file.h>
15 #include <linux/slab.h>
16 #include <linux/kmod.h>
17 #include <linux/kobj_map.h>
18 #include <linux/mutex.h>
19 #include <linux/idr.h>
20 #include <linux/log2.h>
21 #include <linux/pm_runtime.h>
25 static DEFINE_MUTEX(block_class_lock);
26 struct kobject *block_depr;
28 /* for extended dynamic devt allocation, currently only one major is used */
29 #define NR_EXT_DEVT (1 << MINORBITS)
31 /* For extended devt allocation. ext_devt_lock prevents look up
32 * results from going away underneath its user.
34 static DEFINE_SPINLOCK(ext_devt_lock);
35 static DEFINE_IDR(ext_devt_idr);
37 static struct device_type disk_type;
39 static void disk_check_events(struct disk_events *ev,
40 unsigned int *clearing_ptr);
41 static void disk_alloc_events(struct gendisk *disk);
42 static void disk_add_events(struct gendisk *disk);
43 static void disk_del_events(struct gendisk *disk);
44 static void disk_release_events(struct gendisk *disk);
47 * disk_get_part - get partition
48 * @disk: disk to look partition from
49 * @partno: partition number
51 * Look for partition @partno from @disk. If found, increment
52 * reference count and return it.
58 * Pointer to the found partition on success, NULL if not found.
60 struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
62 struct hd_struct *part = NULL;
63 struct disk_part_tbl *ptbl;
65 if (unlikely(partno < 0))
70 ptbl = rcu_dereference(disk->part_tbl);
71 if (likely(partno < ptbl->len)) {
72 part = rcu_dereference(ptbl->part[partno]);
74 get_device(part_to_dev(part));
81 EXPORT_SYMBOL_GPL(disk_get_part);
84 * disk_part_iter_init - initialize partition iterator
85 * @piter: iterator to initialize
86 * @disk: disk to iterate over
87 * @flags: DISK_PITER_* flags
89 * Initialize @piter so that it iterates over partitions of @disk.
94 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
97 struct disk_part_tbl *ptbl;
100 ptbl = rcu_dereference(disk->part_tbl);
105 if (flags & DISK_PITER_REVERSE)
106 piter->idx = ptbl->len - 1;
107 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
112 piter->flags = flags;
116 EXPORT_SYMBOL_GPL(disk_part_iter_init);
119 * disk_part_iter_next - proceed iterator to the next partition and return it
120 * @piter: iterator of interest
122 * Proceed @piter to the next partition and return it.
127 struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
129 struct disk_part_tbl *ptbl;
132 /* put the last partition */
133 disk_put_part(piter->part);
138 ptbl = rcu_dereference(piter->disk->part_tbl);
140 /* determine iteration parameters */
141 if (piter->flags & DISK_PITER_REVERSE) {
143 if (piter->flags & (DISK_PITER_INCL_PART0 |
144 DISK_PITER_INCL_EMPTY_PART0))
153 /* iterate to the next partition */
154 for (; piter->idx != end; piter->idx += inc) {
155 struct hd_struct *part;
157 part = rcu_dereference(ptbl->part[piter->idx]);
160 if (!part_nr_sects_read(part) &&
161 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
162 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
166 get_device(part_to_dev(part));
176 EXPORT_SYMBOL_GPL(disk_part_iter_next);
179 * disk_part_iter_exit - finish up partition iteration
180 * @piter: iter of interest
182 * Called when iteration is over. Cleans up @piter.
187 void disk_part_iter_exit(struct disk_part_iter *piter)
189 disk_put_part(piter->part);
192 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
194 static inline int sector_in_part(struct hd_struct *part, sector_t sector)
196 return part->start_sect <= sector &&
197 sector < part->start_sect + part_nr_sects_read(part);
201 * disk_map_sector_rcu - map sector to partition
202 * @disk: gendisk of interest
203 * @sector: sector to map
205 * Find out which partition @sector maps to on @disk. This is
206 * primarily used for stats accounting.
209 * RCU read locked. The returned partition pointer is valid only
210 * while preemption is disabled.
213 * Found partition on success, part0 is returned if no partition matches
215 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
217 struct disk_part_tbl *ptbl;
218 struct hd_struct *part;
221 ptbl = rcu_dereference(disk->part_tbl);
223 part = rcu_dereference(ptbl->last_lookup);
224 if (part && sector_in_part(part, sector))
227 for (i = 1; i < ptbl->len; i++) {
228 part = rcu_dereference(ptbl->part[i]);
230 if (part && sector_in_part(part, sector)) {
231 rcu_assign_pointer(ptbl->last_lookup, part);
237 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
240 * Can be deleted altogether. Later.
243 static struct blk_major_name {
244 struct blk_major_name *next;
247 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
249 /* index in the above - for now: assume no multimajor ranges */
250 static inline int major_to_index(unsigned major)
252 return major % BLKDEV_MAJOR_HASH_SIZE;
255 #ifdef CONFIG_PROC_FS
256 void blkdev_show(struct seq_file *seqf, off_t offset)
258 struct blk_major_name *dp;
260 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
261 mutex_lock(&block_class_lock);
262 for (dp = major_names[offset]; dp; dp = dp->next)
263 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
264 mutex_unlock(&block_class_lock);
267 #endif /* CONFIG_PROC_FS */
270 * register_blkdev - register a new block device
272 * @major: the requested major device number [1..255]. If @major=0, try to
273 * allocate any unused major number.
274 * @name: the name of the new block device as a zero terminated string
276 * The @name must be unique within the system.
278 * The return value depends on the @major input parameter.
279 * - if a major device number was requested in range [1..255] then the
280 * function returns zero on success, or a negative error code
281 * - if any unused major number was requested with @major=0 parameter
282 * then the return value is the allocated major number in range
283 * [1..255] or a negative error code otherwise
285 int register_blkdev(unsigned int major, const char *name)
287 struct blk_major_name **n, *p;
290 mutex_lock(&block_class_lock);
294 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
295 if (major_names[index] == NULL)
300 printk("register_blkdev: failed to get major for %s\n",
309 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
316 strlcpy(p->name, name, sizeof(p->name));
318 index = major_to_index(major);
320 for (n = &major_names[index]; *n; n = &(*n)->next) {
321 if ((*n)->major == major)
330 printk("register_blkdev: cannot get major %d for %s\n",
335 mutex_unlock(&block_class_lock);
339 EXPORT_SYMBOL(register_blkdev);
341 void unregister_blkdev(unsigned int major, const char *name)
343 struct blk_major_name **n;
344 struct blk_major_name *p = NULL;
345 int index = major_to_index(major);
347 mutex_lock(&block_class_lock);
348 for (n = &major_names[index]; *n; n = &(*n)->next)
349 if ((*n)->major == major)
351 if (!*n || strcmp((*n)->name, name)) {
357 mutex_unlock(&block_class_lock);
361 EXPORT_SYMBOL(unregister_blkdev);
363 static struct kobj_map *bdev_map;
366 * blk_mangle_minor - scatter minor numbers apart
367 * @minor: minor number to mangle
369 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
370 * is enabled. Mangling twice gives the original value.
378 static int blk_mangle_minor(int minor)
380 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
383 for (i = 0; i < MINORBITS / 2; i++) {
384 int low = minor & (1 << i);
385 int high = minor & (1 << (MINORBITS - 1 - i));
386 int distance = MINORBITS - 1 - 2 * i;
388 minor ^= low | high; /* clear both bits */
389 low <<= distance; /* swap the positions */
391 minor |= low | high; /* and set */
398 * blk_alloc_devt - allocate a dev_t for a partition
399 * @part: partition to allocate dev_t for
400 * @devt: out parameter for resulting dev_t
402 * Allocate a dev_t for block device.
405 * 0 on success, allocated dev_t is returned in *@devt. -errno on
411 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
413 struct gendisk *disk = part_to_disk(part);
416 /* in consecutive minor range? */
417 if (part->partno < disk->minors) {
418 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
422 /* allocate ext devt */
423 idr_preload(GFP_KERNEL);
425 spin_lock(&ext_devt_lock);
426 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
427 spin_unlock(&ext_devt_lock);
431 return idx == -ENOSPC ? -EBUSY : idx;
433 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
438 * blk_free_devt - free a dev_t
439 * @devt: dev_t to free
441 * Free @devt which was allocated using blk_alloc_devt().
446 void blk_free_devt(dev_t devt)
448 if (devt == MKDEV(0, 0))
451 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
452 spin_lock(&ext_devt_lock);
453 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
454 spin_unlock(&ext_devt_lock);
458 static char *bdevt_str(dev_t devt, char *buf)
460 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
461 char tbuf[BDEVT_SIZE];
462 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
463 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
465 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
471 * Register device numbers dev..(dev+range-1)
472 * range must be nonzero
473 * The hash chain is sorted on range, so that subranges can override.
475 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
476 struct kobject *(*probe)(dev_t, int *, void *),
477 int (*lock)(dev_t, void *), void *data)
479 kobj_map(bdev_map, devt, range, module, probe, lock, data);
482 EXPORT_SYMBOL(blk_register_region);
484 void blk_unregister_region(dev_t devt, unsigned long range)
486 kobj_unmap(bdev_map, devt, range);
489 EXPORT_SYMBOL(blk_unregister_region);
491 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
493 struct gendisk *p = data;
495 return &disk_to_dev(p)->kobj;
498 static int exact_lock(dev_t devt, void *data)
500 struct gendisk *p = data;
507 static void register_disk(struct gendisk *disk)
509 struct device *ddev = disk_to_dev(disk);
510 struct block_device *bdev;
511 struct disk_part_iter piter;
512 struct hd_struct *part;
515 ddev->parent = disk->driverfs_dev;
517 dev_set_name(ddev, "%s", disk->disk_name);
519 /* delay uevents, until we scanned partition table */
520 dev_set_uevent_suppress(ddev, 1);
522 if (device_add(ddev))
524 if (!sysfs_deprecated) {
525 err = sysfs_create_link(block_depr, &ddev->kobj,
526 kobject_name(&ddev->kobj));
534 * avoid probable deadlock caused by allocating memory with
535 * GFP_KERNEL in runtime_resume callback of its all ancestor
538 pm_runtime_set_memalloc_noio(ddev, true);
540 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
541 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
543 /* No minors to use for partitions */
544 if (!disk_part_scan_enabled(disk))
547 /* No such device (e.g., media were just removed) */
548 if (!get_capacity(disk))
551 bdev = bdget_disk(disk, 0);
555 bdev->bd_invalidated = 1;
556 err = blkdev_get(bdev, FMODE_READ, NULL);
559 blkdev_put(bdev, FMODE_READ);
562 /* announce disk after possible partitions are created */
563 dev_set_uevent_suppress(ddev, 0);
564 kobject_uevent(&ddev->kobj, KOBJ_ADD);
566 /* announce possible partitions */
567 disk_part_iter_init(&piter, disk, 0);
568 while ((part = disk_part_iter_next(&piter)))
569 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
570 disk_part_iter_exit(&piter);
574 * add_disk - add partitioning information to kernel list
575 * @disk: per-device partitioning information
577 * This function registers the partitioning information in @disk
580 * FIXME: error handling
582 void add_disk(struct gendisk *disk)
584 struct backing_dev_info *bdi;
588 /* minors == 0 indicates to use ext devt from part0 and should
589 * be accompanied with EXT_DEVT flag. Make sure all
590 * parameters make sense.
592 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
593 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
595 disk->flags |= GENHD_FL_UP;
597 retval = blk_alloc_devt(&disk->part0, &devt);
602 disk_to_dev(disk)->devt = devt;
604 /* ->major and ->first_minor aren't supposed to be
605 * dereferenced from here on, but set them just in case.
607 disk->major = MAJOR(devt);
608 disk->first_minor = MINOR(devt);
610 disk_alloc_events(disk);
612 /* Register BDI before referencing it from bdev */
613 bdi = &disk->queue->backing_dev_info;
614 bdi_register_dev(bdi, disk_devt(disk));
616 blk_register_region(disk_devt(disk), disk->minors, NULL,
617 exact_match, exact_lock, disk);
619 blk_register_queue(disk);
622 * Take an extra ref on queue which will be put on disk_release()
623 * so that it sticks around as long as @disk is there.
625 WARN_ON_ONCE(!blk_get_queue(disk->queue));
627 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
631 disk_add_events(disk);
633 EXPORT_SYMBOL(add_disk);
635 void del_gendisk(struct gendisk *disk)
637 struct disk_part_iter piter;
638 struct hd_struct *part;
640 disk_del_events(disk);
642 /* invalidate stuff */
643 disk_part_iter_init(&piter, disk,
644 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
645 while ((part = disk_part_iter_next(&piter))) {
646 invalidate_partition(disk, part->partno);
647 delete_partition(disk, part->partno);
649 disk_part_iter_exit(&piter);
651 invalidate_partition(disk, 0);
652 set_capacity(disk, 0);
653 disk->flags &= ~GENHD_FL_UP;
655 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
656 bdi_unregister(&disk->queue->backing_dev_info);
657 blk_unregister_queue(disk);
658 blk_unregister_region(disk_devt(disk), disk->minors);
660 part_stat_set_all(&disk->part0, 0);
661 disk->part0.stamp = 0;
663 kobject_put(disk->part0.holder_dir);
664 kobject_put(disk->slave_dir);
665 disk->driverfs_dev = NULL;
666 if (!sysfs_deprecated)
667 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
668 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
669 device_del(disk_to_dev(disk));
671 EXPORT_SYMBOL(del_gendisk);
674 * get_gendisk - get partitioning information for a given device
675 * @devt: device to get partitioning information for
676 * @partno: returned partition index
678 * This function gets the structure containing partitioning
679 * information for the given device @devt.
681 struct gendisk *get_gendisk(dev_t devt, int *partno)
683 struct gendisk *disk = NULL;
685 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
686 struct kobject *kobj;
688 kobj = kobj_lookup(bdev_map, devt, partno);
690 disk = dev_to_disk(kobj_to_dev(kobj));
692 struct hd_struct *part;
694 spin_lock(&ext_devt_lock);
695 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
696 if (part && get_disk(part_to_disk(part))) {
697 *partno = part->partno;
698 disk = part_to_disk(part);
700 spin_unlock(&ext_devt_lock);
705 EXPORT_SYMBOL(get_gendisk);
708 * bdget_disk - do bdget() by gendisk and partition number
709 * @disk: gendisk of interest
710 * @partno: partition number
712 * Find partition @partno from @disk, do bdget() on it.
718 * Resulting block_device on success, NULL on failure.
720 struct block_device *bdget_disk(struct gendisk *disk, int partno)
722 struct hd_struct *part;
723 struct block_device *bdev = NULL;
725 part = disk_get_part(disk, partno);
727 bdev = bdget(part_devt(part));
732 EXPORT_SYMBOL(bdget_disk);
735 * print a full list of all partitions - intended for places where the root
736 * filesystem can't be mounted and thus to give the victim some idea of what
739 void __init printk_all_partitions(void)
741 struct class_dev_iter iter;
744 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
745 while ((dev = class_dev_iter_next(&iter))) {
746 struct gendisk *disk = dev_to_disk(dev);
747 struct disk_part_iter piter;
748 struct hd_struct *part;
749 char name_buf[BDEVNAME_SIZE];
750 char devt_buf[BDEVT_SIZE];
753 * Don't show empty devices or things that have been
756 if (get_capacity(disk) == 0 ||
757 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
761 * Note, unlike /proc/partitions, I am showing the
762 * numbers in hex - the same format as the root=
765 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
766 while ((part = disk_part_iter_next(&piter))) {
767 bool is_part0 = part == &disk->part0;
769 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
770 bdevt_str(part_devt(part), devt_buf),
771 (unsigned long long)part_nr_sects_read(part) >> 1
772 , disk_name(disk, part->partno, name_buf),
773 part->info ? part->info->uuid : "");
775 if (disk->driverfs_dev != NULL &&
776 disk->driverfs_dev->driver != NULL)
777 printk(" driver: %s\n",
778 disk->driverfs_dev->driver->name);
780 printk(" (driver?)\n");
784 disk_part_iter_exit(&piter);
786 class_dev_iter_exit(&iter);
789 #ifdef CONFIG_PROC_FS
791 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
794 struct class_dev_iter *iter;
797 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
799 return ERR_PTR(-ENOMEM);
801 seqf->private = iter;
802 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
804 dev = class_dev_iter_next(iter);
809 return dev_to_disk(dev);
812 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
817 dev = class_dev_iter_next(seqf->private);
819 return dev_to_disk(dev);
824 static void disk_seqf_stop(struct seq_file *seqf, void *v)
826 struct class_dev_iter *iter = seqf->private;
828 /* stop is called even after start failed :-( */
830 class_dev_iter_exit(iter);
835 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
839 p = disk_seqf_start(seqf, pos);
840 if (!IS_ERR_OR_NULL(p) && !*pos)
841 seq_puts(seqf, "major minor #blocks name\n\n");
845 static int show_partition(struct seq_file *seqf, void *v)
847 struct gendisk *sgp = v;
848 struct disk_part_iter piter;
849 struct hd_struct *part;
850 char buf[BDEVNAME_SIZE];
852 /* Don't show non-partitionable removeable devices or empty devices */
853 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
854 (sgp->flags & GENHD_FL_REMOVABLE)))
856 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
859 /* show the full disk and all non-0 size partitions of it */
860 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
861 while ((part = disk_part_iter_next(&piter)))
862 seq_printf(seqf, "%4d %7d %10llu %s\n",
863 MAJOR(part_devt(part)), MINOR(part_devt(part)),
864 (unsigned long long)part_nr_sects_read(part) >> 1,
865 disk_name(sgp, part->partno, buf));
866 disk_part_iter_exit(&piter);
871 static const struct seq_operations partitions_op = {
872 .start = show_partition_start,
873 .next = disk_seqf_next,
874 .stop = disk_seqf_stop,
875 .show = show_partition
878 static int partitions_open(struct inode *inode, struct file *file)
880 return seq_open(file, &partitions_op);
883 static const struct file_operations proc_partitions_operations = {
884 .open = partitions_open,
887 .release = seq_release,
892 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
894 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
895 /* Make old-style 2.4 aliases work */
896 request_module("block-major-%d", MAJOR(devt));
900 static int __init genhd_device_init(void)
904 block_class.dev_kobj = sysfs_dev_block_kobj;
905 error = class_register(&block_class);
908 bdev_map = kobj_map_init(base_probe, &block_class_lock);
911 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
913 /* create top-level block dir */
914 if (!sysfs_deprecated)
915 block_depr = kobject_create_and_add("block", NULL);
919 subsys_initcall(genhd_device_init);
921 static ssize_t disk_range_show(struct device *dev,
922 struct device_attribute *attr, char *buf)
924 struct gendisk *disk = dev_to_disk(dev);
926 return sprintf(buf, "%d\n", disk->minors);
929 static ssize_t disk_ext_range_show(struct device *dev,
930 struct device_attribute *attr, char *buf)
932 struct gendisk *disk = dev_to_disk(dev);
934 return sprintf(buf, "%d\n", disk_max_parts(disk));
937 static ssize_t disk_removable_show(struct device *dev,
938 struct device_attribute *attr, char *buf)
940 struct gendisk *disk = dev_to_disk(dev);
942 return sprintf(buf, "%d\n",
943 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
946 static ssize_t disk_ro_show(struct device *dev,
947 struct device_attribute *attr, char *buf)
949 struct gendisk *disk = dev_to_disk(dev);
951 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
954 static ssize_t disk_capability_show(struct device *dev,
955 struct device_attribute *attr, char *buf)
957 struct gendisk *disk = dev_to_disk(dev);
959 return sprintf(buf, "%x\n", disk->flags);
962 static ssize_t disk_alignment_offset_show(struct device *dev,
963 struct device_attribute *attr,
966 struct gendisk *disk = dev_to_disk(dev);
968 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
971 static ssize_t disk_discard_alignment_show(struct device *dev,
972 struct device_attribute *attr,
975 struct gendisk *disk = dev_to_disk(dev);
977 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
980 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
981 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
982 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
983 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
984 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
985 static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
986 static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
988 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
989 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
990 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
991 #ifdef CONFIG_FAIL_MAKE_REQUEST
992 static struct device_attribute dev_attr_fail =
993 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
995 #ifdef CONFIG_FAIL_IO_TIMEOUT
996 static struct device_attribute dev_attr_fail_timeout =
997 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
1001 static struct attribute *disk_attrs[] = {
1002 &dev_attr_range.attr,
1003 &dev_attr_ext_range.attr,
1004 &dev_attr_removable.attr,
1006 &dev_attr_size.attr,
1007 &dev_attr_alignment_offset.attr,
1008 &dev_attr_discard_alignment.attr,
1009 &dev_attr_capability.attr,
1010 &dev_attr_stat.attr,
1011 &dev_attr_inflight.attr,
1012 #ifdef CONFIG_FAIL_MAKE_REQUEST
1013 &dev_attr_fail.attr,
1015 #ifdef CONFIG_FAIL_IO_TIMEOUT
1016 &dev_attr_fail_timeout.attr,
1021 static struct attribute_group disk_attr_group = {
1022 .attrs = disk_attrs,
1025 static const struct attribute_group *disk_attr_groups[] = {
1031 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1032 * @disk: disk to replace part_tbl for
1033 * @new_ptbl: new part_tbl to install
1035 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1036 * original ptbl is freed using RCU callback.
1039 * Matching bd_mutx locked.
1041 static void disk_replace_part_tbl(struct gendisk *disk,
1042 struct disk_part_tbl *new_ptbl)
1044 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1046 rcu_assign_pointer(disk->part_tbl, new_ptbl);
1049 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
1050 kfree_rcu(old_ptbl, rcu_head);
1055 * disk_expand_part_tbl - expand disk->part_tbl
1056 * @disk: disk to expand part_tbl for
1057 * @partno: expand such that this partno can fit in
1059 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1060 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1063 * Matching bd_mutex locked, might sleep.
1066 * 0 on success, -errno on failure.
1068 int disk_expand_part_tbl(struct gendisk *disk, int partno)
1070 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1071 struct disk_part_tbl *new_ptbl;
1072 int len = old_ptbl ? old_ptbl->len : 0;
1077 * check for int overflow, since we can get here from blkpg_ioctl()
1078 * with a user passed 'partno'.
1080 target = partno + 1;
1084 /* disk_max_parts() is zero during initialization, ignore if so */
1085 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1091 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1092 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1096 new_ptbl->len = target;
1098 for (i = 0; i < len; i++)
1099 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1101 disk_replace_part_tbl(disk, new_ptbl);
1105 static void disk_release(struct device *dev)
1107 struct gendisk *disk = dev_to_disk(dev);
1109 blk_free_devt(dev->devt);
1110 disk_release_events(disk);
1111 kfree(disk->random);
1112 disk_replace_part_tbl(disk, NULL);
1113 free_part_stats(&disk->part0);
1114 free_part_info(&disk->part0);
1116 blk_put_queue(disk->queue);
1120 static int disk_uevent(struct device *dev, struct kobj_uevent_env *env)
1122 struct gendisk *disk = dev_to_disk(dev);
1123 struct disk_part_iter piter;
1124 struct hd_struct *part;
1127 disk_part_iter_init(&piter, disk, 0);
1128 while((part = disk_part_iter_next(&piter)))
1130 disk_part_iter_exit(&piter);
1131 add_uevent_var(env, "NPARTS=%u", cnt);
1135 struct class block_class = {
1139 static char *block_devnode(struct device *dev, umode_t *mode,
1140 kuid_t *uid, kgid_t *gid)
1142 struct gendisk *disk = dev_to_disk(dev);
1145 return disk->devnode(disk, mode);
1149 static struct device_type disk_type = {
1151 .groups = disk_attr_groups,
1152 .release = disk_release,
1153 .devnode = block_devnode,
1154 .uevent = disk_uevent,
1157 #ifdef CONFIG_PROC_FS
1159 * aggregate disk stat collector. Uses the same stats that the sysfs
1160 * entries do, above, but makes them available through one seq_file.
1162 * The output looks suspiciously like /proc/partitions with a bunch of
1165 static int diskstats_show(struct seq_file *seqf, void *v)
1167 struct gendisk *gp = v;
1168 struct disk_part_iter piter;
1169 struct hd_struct *hd;
1170 char buf[BDEVNAME_SIZE];
1174 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1175 seq_puts(seqf, "major minor name"
1176 " rio rmerge rsect ruse wio wmerge "
1177 "wsect wuse running use aveq"
1181 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
1182 while ((hd = disk_part_iter_next(&piter))) {
1183 cpu = part_stat_lock();
1184 part_round_stats(cpu, hd);
1186 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1187 "%u %lu %lu %lu %u %u %u %u\n",
1188 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1189 disk_name(gp, hd->partno, buf),
1190 part_stat_read(hd, ios[READ]),
1191 part_stat_read(hd, merges[READ]),
1192 part_stat_read(hd, sectors[READ]),
1193 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1194 part_stat_read(hd, ios[WRITE]),
1195 part_stat_read(hd, merges[WRITE]),
1196 part_stat_read(hd, sectors[WRITE]),
1197 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
1199 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1200 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1203 disk_part_iter_exit(&piter);
1208 static const struct seq_operations diskstats_op = {
1209 .start = disk_seqf_start,
1210 .next = disk_seqf_next,
1211 .stop = disk_seqf_stop,
1212 .show = diskstats_show
1215 static int diskstats_open(struct inode *inode, struct file *file)
1217 return seq_open(file, &diskstats_op);
1220 static const struct file_operations proc_diskstats_operations = {
1221 .open = diskstats_open,
1223 .llseek = seq_lseek,
1224 .release = seq_release,
1227 static int __init proc_genhd_init(void)
1229 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1230 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1233 module_init(proc_genhd_init);
1234 #endif /* CONFIG_PROC_FS */
1236 dev_t blk_lookup_devt(const char *name, int partno)
1238 dev_t devt = MKDEV(0, 0);
1239 struct class_dev_iter iter;
1242 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1243 while ((dev = class_dev_iter_next(&iter))) {
1244 struct gendisk *disk = dev_to_disk(dev);
1245 struct hd_struct *part;
1247 if (strcmp(dev_name(dev), name))
1250 if (partno < disk->minors) {
1251 /* We need to return the right devno, even
1252 * if the partition doesn't exist yet.
1254 devt = MKDEV(MAJOR(dev->devt),
1255 MINOR(dev->devt) + partno);
1258 part = disk_get_part(disk, partno);
1260 devt = part_devt(part);
1261 disk_put_part(part);
1264 disk_put_part(part);
1266 class_dev_iter_exit(&iter);
1269 EXPORT_SYMBOL(blk_lookup_devt);
1271 struct gendisk *alloc_disk(int minors)
1273 return alloc_disk_node(minors, NUMA_NO_NODE);
1275 EXPORT_SYMBOL(alloc_disk);
1277 struct gendisk *alloc_disk_node(int minors, int node_id)
1279 struct gendisk *disk;
1281 disk = kmalloc_node(sizeof(struct gendisk),
1282 GFP_KERNEL | __GFP_ZERO, node_id);
1284 if (!init_part_stats(&disk->part0)) {
1288 disk->node_id = node_id;
1289 if (disk_expand_part_tbl(disk, 0)) {
1290 free_part_stats(&disk->part0);
1294 disk->part_tbl->part[0] = &disk->part0;
1297 * set_capacity() and get_capacity() currently don't use
1298 * seqcounter to read/update the part0->nr_sects. Still init
1299 * the counter as we can read the sectors in IO submission
1300 * patch using seqence counters.
1302 * TODO: Ideally set_capacity() and get_capacity() should be
1303 * converted to make use of bd_mutex and sequence counters.
1305 seqcount_init(&disk->part0.nr_sects_seq);
1306 hd_ref_init(&disk->part0);
1308 disk->minors = minors;
1309 rand_initialize_disk(disk);
1310 disk_to_dev(disk)->class = &block_class;
1311 disk_to_dev(disk)->type = &disk_type;
1312 device_initialize(disk_to_dev(disk));
1316 EXPORT_SYMBOL(alloc_disk_node);
1318 struct kobject *get_disk(struct gendisk *disk)
1320 struct module *owner;
1321 struct kobject *kobj;
1325 owner = disk->fops->owner;
1326 if (owner && !try_module_get(owner))
1328 kobj = kobject_get(&disk_to_dev(disk)->kobj);
1337 EXPORT_SYMBOL(get_disk);
1339 void put_disk(struct gendisk *disk)
1342 kobject_put(&disk_to_dev(disk)->kobj);
1345 EXPORT_SYMBOL(put_disk);
1347 static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1349 char event[] = "DISK_RO=1";
1350 char *envp[] = { event, NULL };
1354 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1357 void set_device_ro(struct block_device *bdev, int flag)
1359 bdev->bd_part->policy = flag;
1362 EXPORT_SYMBOL(set_device_ro);
1364 void set_disk_ro(struct gendisk *disk, int flag)
1366 struct disk_part_iter piter;
1367 struct hd_struct *part;
1369 if (disk->part0.policy != flag) {
1370 set_disk_ro_uevent(disk, flag);
1371 disk->part0.policy = flag;
1374 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
1375 while ((part = disk_part_iter_next(&piter)))
1376 part->policy = flag;
1377 disk_part_iter_exit(&piter);
1380 EXPORT_SYMBOL(set_disk_ro);
1382 int bdev_read_only(struct block_device *bdev)
1386 return bdev->bd_part->policy;
1389 EXPORT_SYMBOL(bdev_read_only);
1391 int invalidate_partition(struct gendisk *disk, int partno)
1394 struct block_device *bdev = bdget_disk(disk, partno);
1397 res = __invalidate_device(bdev, true);
1403 EXPORT_SYMBOL(invalidate_partition);
1406 * Disk events - monitor disk events like media change and eject request.
1408 struct disk_events {
1409 struct list_head node; /* all disk_event's */
1410 struct gendisk *disk; /* the associated disk */
1413 struct mutex block_mutex; /* protects blocking */
1414 int block; /* event blocking depth */
1415 unsigned int pending; /* events already sent out */
1416 unsigned int clearing; /* events being cleared */
1418 long poll_msecs; /* interval, -1 for default */
1419 struct delayed_work dwork;
1422 static const char *disk_events_strs[] = {
1423 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1424 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1427 static char *disk_uevents[] = {
1428 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1429 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1432 /* list of all disk_events */
1433 static DEFINE_MUTEX(disk_events_mutex);
1434 static LIST_HEAD(disk_events);
1436 /* disable in-kernel polling by default */
1437 static unsigned long disk_events_dfl_poll_msecs = 0;
1439 static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1441 struct disk_events *ev = disk->ev;
1442 long intv_msecs = 0;
1445 * If device-specific poll interval is set, always use it. If
1446 * the default is being used, poll iff there are events which
1447 * can't be monitored asynchronously.
1449 if (ev->poll_msecs >= 0)
1450 intv_msecs = ev->poll_msecs;
1451 else if (disk->events & ~disk->async_events)
1452 intv_msecs = disk_events_dfl_poll_msecs;
1454 return msecs_to_jiffies(intv_msecs);
1458 * disk_block_events - block and flush disk event checking
1459 * @disk: disk to block events for
1461 * On return from this function, it is guaranteed that event checking
1462 * isn't in progress and won't happen until unblocked by
1463 * disk_unblock_events(). Events blocking is counted and the actual
1464 * unblocking happens after the matching number of unblocks are done.
1466 * Note that this intentionally does not block event checking from
1467 * disk_clear_events().
1472 void disk_block_events(struct gendisk *disk)
1474 struct disk_events *ev = disk->ev;
1475 unsigned long flags;
1482 * Outer mutex ensures that the first blocker completes canceling
1483 * the event work before further blockers are allowed to finish.
1485 mutex_lock(&ev->block_mutex);
1487 spin_lock_irqsave(&ev->lock, flags);
1488 cancel = !ev->block++;
1489 spin_unlock_irqrestore(&ev->lock, flags);
1492 cancel_delayed_work_sync(&disk->ev->dwork);
1494 mutex_unlock(&ev->block_mutex);
1497 static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1499 struct disk_events *ev = disk->ev;
1501 unsigned long flags;
1503 spin_lock_irqsave(&ev->lock, flags);
1505 if (WARN_ON_ONCE(ev->block <= 0))
1512 * Not exactly a latency critical operation, set poll timer
1513 * slack to 25% and kick event check.
1515 intv = disk_events_poll_jiffies(disk);
1516 set_timer_slack(&ev->dwork.timer, intv / 4);
1518 queue_delayed_work(system_freezable_power_efficient_wq,
1521 queue_delayed_work(system_freezable_power_efficient_wq,
1524 spin_unlock_irqrestore(&ev->lock, flags);
1528 * disk_unblock_events - unblock disk event checking
1529 * @disk: disk to unblock events for
1531 * Undo disk_block_events(). When the block count reaches zero, it
1532 * starts events polling if configured.
1535 * Don't care. Safe to call from irq context.
1537 void disk_unblock_events(struct gendisk *disk)
1540 __disk_unblock_events(disk, false);
1544 * disk_flush_events - schedule immediate event checking and flushing
1545 * @disk: disk to check and flush events for
1546 * @mask: events to flush
1548 * Schedule immediate event checking on @disk if not blocked. Events in
1549 * @mask are scheduled to be cleared from the driver. Note that this
1550 * doesn't clear the events from @disk->ev.
1553 * If @mask is non-zero must be called with bdev->bd_mutex held.
1555 void disk_flush_events(struct gendisk *disk, unsigned int mask)
1557 struct disk_events *ev = disk->ev;
1562 spin_lock_irq(&ev->lock);
1563 ev->clearing |= mask;
1565 mod_delayed_work(system_freezable_power_efficient_wq,
1567 spin_unlock_irq(&ev->lock);
1571 * disk_clear_events - synchronously check, clear and return pending events
1572 * @disk: disk to fetch and clear events from
1573 * @mask: mask of events to be fetched and clearted
1575 * Disk events are synchronously checked and pending events in @mask
1576 * are cleared and returned. This ignores the block count.
1581 unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1583 const struct block_device_operations *bdops = disk->fops;
1584 struct disk_events *ev = disk->ev;
1585 unsigned int pending;
1586 unsigned int clearing = mask;
1589 /* for drivers still using the old ->media_changed method */
1590 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1591 bdops->media_changed && bdops->media_changed(disk))
1592 return DISK_EVENT_MEDIA_CHANGE;
1596 disk_block_events(disk);
1599 * store the union of mask and ev->clearing on the stack so that the
1600 * race with disk_flush_events does not cause ambiguity (ev->clearing
1601 * can still be modified even if events are blocked).
1603 spin_lock_irq(&ev->lock);
1604 clearing |= ev->clearing;
1606 spin_unlock_irq(&ev->lock);
1608 disk_check_events(ev, &clearing);
1610 * if ev->clearing is not 0, the disk_flush_events got called in the
1611 * middle of this function, so we want to run the workfn without delay.
1613 __disk_unblock_events(disk, ev->clearing ? true : false);
1615 /* then, fetch and clear pending events */
1616 spin_lock_irq(&ev->lock);
1617 pending = ev->pending & mask;
1618 ev->pending &= ~mask;
1619 spin_unlock_irq(&ev->lock);
1620 WARN_ON_ONCE(clearing & mask);
1626 * Separate this part out so that a different pointer for clearing_ptr can be
1627 * passed in for disk_clear_events.
1629 static void disk_events_workfn(struct work_struct *work)
1631 struct delayed_work *dwork = to_delayed_work(work);
1632 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1634 disk_check_events(ev, &ev->clearing);
1637 static void disk_check_events(struct disk_events *ev,
1638 unsigned int *clearing_ptr)
1640 struct gendisk *disk = ev->disk;
1641 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1642 unsigned int clearing = *clearing_ptr;
1643 unsigned int events;
1645 int nr_events = 0, i;
1648 events = disk->fops->check_events(disk, clearing);
1650 /* accumulate pending events and schedule next poll if necessary */
1651 spin_lock_irq(&ev->lock);
1653 events &= ~ev->pending;
1654 ev->pending |= events;
1655 *clearing_ptr &= ~clearing;
1657 intv = disk_events_poll_jiffies(disk);
1658 if (!ev->block && intv)
1659 queue_delayed_work(system_freezable_power_efficient_wq,
1662 spin_unlock_irq(&ev->lock);
1665 * Tell userland about new events. Only the events listed in
1666 * @disk->events are reported. Unlisted events are processed the
1667 * same internally but never get reported to userland.
1669 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
1670 if (events & disk->events & (1 << i))
1671 envp[nr_events++] = disk_uevents[i];
1674 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1678 * A disk events enabled device has the following sysfs nodes under
1679 * its /sys/block/X/ directory.
1681 * events : list of all supported events
1682 * events_async : list of events which can be detected w/o polling
1683 * events_poll_msecs : polling interval, 0: disable, -1: system default
1685 static ssize_t __disk_events_show(unsigned int events, char *buf)
1687 const char *delim = "";
1691 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1692 if (events & (1 << i)) {
1693 pos += sprintf(buf + pos, "%s%s",
1694 delim, disk_events_strs[i]);
1698 pos += sprintf(buf + pos, "\n");
1702 static ssize_t disk_events_show(struct device *dev,
1703 struct device_attribute *attr, char *buf)
1705 struct gendisk *disk = dev_to_disk(dev);
1707 return __disk_events_show(disk->events, buf);
1710 static ssize_t disk_events_async_show(struct device *dev,
1711 struct device_attribute *attr, char *buf)
1713 struct gendisk *disk = dev_to_disk(dev);
1715 return __disk_events_show(disk->async_events, buf);
1718 static ssize_t disk_events_poll_msecs_show(struct device *dev,
1719 struct device_attribute *attr,
1722 struct gendisk *disk = dev_to_disk(dev);
1724 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1727 static ssize_t disk_events_poll_msecs_store(struct device *dev,
1728 struct device_attribute *attr,
1729 const char *buf, size_t count)
1731 struct gendisk *disk = dev_to_disk(dev);
1734 if (!count || !sscanf(buf, "%ld", &intv))
1737 if (intv < 0 && intv != -1)
1740 disk_block_events(disk);
1741 disk->ev->poll_msecs = intv;
1742 __disk_unblock_events(disk, true);
1747 static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1748 static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1749 static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1750 disk_events_poll_msecs_show,
1751 disk_events_poll_msecs_store);
1753 static const struct attribute *disk_events_attrs[] = {
1754 &dev_attr_events.attr,
1755 &dev_attr_events_async.attr,
1756 &dev_attr_events_poll_msecs.attr,
1761 * The default polling interval can be specified by the kernel
1762 * parameter block.events_dfl_poll_msecs which defaults to 0
1763 * (disable). This can also be modified runtime by writing to
1764 * /sys/module/block/events_dfl_poll_msecs.
1766 static int disk_events_set_dfl_poll_msecs(const char *val,
1767 const struct kernel_param *kp)
1769 struct disk_events *ev;
1772 ret = param_set_ulong(val, kp);
1776 mutex_lock(&disk_events_mutex);
1778 list_for_each_entry(ev, &disk_events, node)
1779 disk_flush_events(ev->disk, 0);
1781 mutex_unlock(&disk_events_mutex);
1786 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1787 .set = disk_events_set_dfl_poll_msecs,
1788 .get = param_get_ulong,
1791 #undef MODULE_PARAM_PREFIX
1792 #define MODULE_PARAM_PREFIX "block."
1794 module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1795 &disk_events_dfl_poll_msecs, 0644);
1798 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1800 static void disk_alloc_events(struct gendisk *disk)
1802 struct disk_events *ev;
1804 if (!disk->fops->check_events)
1807 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1809 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1813 INIT_LIST_HEAD(&ev->node);
1815 spin_lock_init(&ev->lock);
1816 mutex_init(&ev->block_mutex);
1818 ev->poll_msecs = -1;
1819 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1824 static void disk_add_events(struct gendisk *disk)
1829 /* FIXME: error handling */
1830 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1831 pr_warn("%s: failed to create sysfs files for events\n",
1834 mutex_lock(&disk_events_mutex);
1835 list_add_tail(&disk->ev->node, &disk_events);
1836 mutex_unlock(&disk_events_mutex);
1839 * Block count is initialized to 1 and the following initial
1840 * unblock kicks it into action.
1842 __disk_unblock_events(disk, true);
1845 static void disk_del_events(struct gendisk *disk)
1850 disk_block_events(disk);
1852 mutex_lock(&disk_events_mutex);
1853 list_del_init(&disk->ev->node);
1854 mutex_unlock(&disk_events_mutex);
1856 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1859 static void disk_release_events(struct gendisk *disk)
1861 /* the block count should be 1 from disk_del_events() */
1862 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);